quantification of nε-(2-furoylmethyl)-l-lysine (furosine), nε ......quantification of...
TRANSCRIPT
Accepted Manuscript
Analytical methods
Quantification of Nε-(2-Furoylmethyl)-L-lysine (furosine), Nε-(Carboxymeth-yl)-L-lysine (CML), Nε-(Carboxyethyl)-L-lysine (CEL) and Total Lysinethrough Stable Isotope Dilution Assay and Tandem Mass Spectrometry
Antonio Dario Troise, Alberto Fiore, Markus Wiltafsky, Vincenzo Fogliano
PII: S0308-8146(15)00707-4DOI: http://dx.doi.org/10.1016/j.foodchem.2015.04.137Reference: FOCH 17538
To appear in: Food Chemistry
Received Date: 14 January 2015Revised Date: 28 April 2015Accepted Date: 29 April 2015
Please cite this article as: Troise, A.D., Fiore, A., Wiltafsky, M., Fogliano, V., Quantification of Nε-(2-Furoylmethyl)-L-lysine (furosine), Nε-(Carboxymethyl)-L-lysine (CML), Nε-(Carboxyethyl)-L-lysine (CEL) andTotal Lysine through Stable Isotope Dilution Assay and Tandem Mass Spectrometry, Food Chemistry (2015), doi:http://dx.doi.org/10.1016/j.foodchem.2015.04.137
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
© 2015, Elsevier. Licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) http://creativecommons.org/licenses/by-nc-nd/4.0/
Quantification of Nε-(2-Furoylmethyl)-L-lysine (furosine), Nε-(Carboxymethyl)-L-lysine
(CML), Nε-(Carboxyethyl)-L-lysine (CEL) and Total Lysine through Stable Isotope Dilution
Assay and Tandem Mass Spectrometry
Antonio Dario Troise1,2*
, Alberto Fiore3, Markus Wiltafsky
4, Vincenzo Fogliano
1
1 Food Quality Design Group, Wageningen University, PO Box 8129, 6700 EV, Wageningen, The
Netherlands
2 Department of Agricultural and Food Science, University of Napoli ―Federico II‖, Parco Gussone,
80055 Portici, Napoli, Italy
3 School of Science, Engineering & Technology Division of Food Science, Abertay University
Dundee DD1 1HG, UK
4Evonik Industries AG, Rodenbacher Chaussee 4, 63457 Hanau, Germany
*Corresponding author
Antonio Dario Troise
Phone number: +39 081 2539360
Abstract
The control of Maillard reaction (MR) is a key point to ensure processed foods quality. Due to the
presence of a primary amino group on its side chain, lysine is particularly prone to chemical
modifications with the formation of Amadori products (AP), Nε-(Carboxymethyl)-L-lysine (CML),
Nε-(Carboxyethyl)-L-lysine (CEL). A new analytical strategy was proposed which allowed to
simultaneously quantify lysine, CML, CEL and the Nε-(2-Furoylmethyl)-L-lysine (furosine), the
indirect marker of AP. The procedure is based on stable isotope dilution assay followed by, liquid
chromatography tandem mass spectrometry. It showed high sensitivity and good reproducibility and
repeatability in different foods. The limit of detection and the RSD% were lower than 5 ppb and
below 8%, respectively. Results obtained with the new procedure not only improved the knowledge
about the reliability of thermal treatment markers, but also defined new insights in the relationship
between Maillard reaction products and their precursors.
Keywords: Maillard reaction, LC-MS/MS, CML, CEL, lysine, furosine
Abbreviations: Maillard reaction (MR), Maillard reaction end products (MRPs) Nε-(2-
Furoylmethyl)-L-lysine (furosine), Nε-(Carboxymethyl)-L-lysine (CML), Nε-(Carboxyethyl)-L-
lysine (CEL),
List of compounds: Nε-(2-Furoylmethyl)-L-lysine (furosine, PubChem CID: no items) , Nε-
(Carboxymethyl)-L-lysine (CML, PubChem CID: 123800), Nε-(Carboxyethyl)-L-lysine (CEL,
PubChem CID: no items), lysine (PubChem CID: 5962).
1. Introduction 1
The final quality of many industrial food products depends on food formulation and processing 2
design resulting in the formation of a huge variety of molecules as a consequence of thermal 3
treatments and chemical changes (van Boekel, Fogliano, Pellegrini, Stanton, Scholz, Lalljie, et al., 4
2010). Along with lipid oxidation, the Maillard reaction (MR) occupies a prominent place in the 5
final quality of food being responsible not only for the desired color and aroma compounds but also 6
for the formation of potentially toxic Maillard reaction end products (MRPs). The reaction between 7
reducing sugars and amino groups is the first step in the Maillard cascade: the formation of the 8
stable 1-amino-1-deoxy-2-ketose the Amadori product (AP) and 2-amino-2-deoxyaldose Heyns 9
products represents the starting point of the many chemical pathways of this reaction (Hodge, 10
1953). The presence of an amino group on the side chain of lysine makes this amino acid 11
particularly sensitive to the carbonyls attachments. The modifications arising from the lysine 12
blockage resulted in the formation of a bewildering array of molecules: Nε-(1-Deoxy-D-fructos-1-13
yl)-L-lysine (fructosyl-lysine), Nε-(Carboxymethyl)-L-lysine (CML), Nε-(Carboxyethyl)-L-lysine 14
(CEL), pentosidine, pyrraline, lysino-alanine, 5-hydroxymethylfurfural (HMF), -dicarbonyls and 15
aroma key odorants (Yaylayan & Huyghuesdespointes, 1994). Fructosyl-lysine, CML and CEL 16
represent the most widely studied MRPs, and they are often used as biomarker of food quality 17
(Erbersdobler & Somoza, 2007; Nguyen, van der Fels-Klerx, & van Boekel, 2013). As highlighted 18
in Figure 1, the acid hydrolysis adopted to release free amino acids from the polypeptide chain 19
promote the conversion of the 1-deoxy-fructosyl-L-Lysine (AP) through a cyclized Schiff base, into 20
the of Nε-(2-furoylmethyl)-L-lysine (furosine) which is a compound that can be quantified after 21
protein hydrolysis and it has been widely used as marker of thermal treatment particularly in the 22
dairy products (Krause, Knoll, & Henle, 2003). 23
The formation of CML and CEL from the oxidation of ARP and HRP has been well characterized 24
(Nguyen, van der Fels-Klerx, & van Boekel, 2013). Carbohydrate fragmentation allows the 25
formation of glyoxal and methylglyoxal that readily react with lysine residues yielding the 26
glycoxidation products CML and CEL, respectively (Ahmed, Thorpe, & Baynes, 1986). Moreover, 27
CML and CEL can be formed via the Namiki-pathway through three subsequent steps: Schiff base 28
production, glycolaldehyde alkylimine synthesis, oxidation and formation of glyoxal or 29
methylglyoxal which react with lysine to yield CML and CEL. Another route of CML and CEL 30
formation is linked to lipid peroxidation as glyoxal and methylglyoxal can derive from 31
polyunsatured fatty acids (Hidalgo & Zamora, 2005). Moreover, the two markers can be also 32
formed from fragmentation and subsequent glycation of ascorbic acid and dehydroascorbic acid 33
(Leclere, Birlouez-Aragon, & Meli, 2002). 34
From the analytical point of view the identification of these markers of heat treatment can be 35
approached in several ways (Tessier & Birlouez-Aragon, 2012). Furosine is used as indirect marker 36
of quality control of moderately heat-treated dairy samples. The golden standards for furosine 37
detection are ion-exchange chromatography, reverse phase high performance liquid 38
chromatography (RP-HPLC) with UV detection, (Henle, Zehetner, & Klostermeyer, 1995) capillary 39
electrophoresis and ion-pairing HPLC by using sodium-heptanosulphonate (Vallejo-Cordoba, 40
Mazorra-Manzano, & Gonzalez-Cordova, 2004). These procedures had several drawbacks mainly 41
related to the modifications occurring during sample preparation: the acidic hydrolysis does not 42
allow the differentiation between AP and glycosyl-amine; overestimation or underestimation linked 43
to the acidic hydrolysis might occur due to the formation of further intermediates and end-products 44
(Pischetsrieder & Henle, 2012). 45
As for furosine, CML and CEL analysis implies acidic hydrolysis to hydrolyze peptide bonds 46
followed by their quantification that could be performed by different instrumental methods 47
(Nguyen, van der Fels-Klerx, & van Boekel, 2013). In some papers a pre-column derivatization 48
with o-phthalaldehyde was used to allow the detection by florescence detector (Hartkopf, Pahlke, 49
Ludemann, & Erbersdobler, 1994), while a widely used approach for CML and CEL detection is 50
gas or liquid chromatography coupled with tandem mass spectrometry. Specifically, multiple 51
reaction monitoring (MRM) mode improves the sensitivity, reduces the coefficient of variability 52
and ruled out the problems of derivatization (Delatour, Hegele, Parisod, Richoz, Maurer, Steven, et 53
al., 2009). A double derivatization is required for GC separation and this bottleneck highlights the 54
advantages of LC-MS/MS detection: no derivatization, highest sensitivity and good reproducibility 55
(Charissou, Ait-Ameur, & Birlouez-Aragon, 2007; Fenaille, Parisod, Visani, Populaire, Tabet, & 56
Guy, 2006). Moreover CML, CEL and lysine detection is possible also by matrix-assisted laser 57
desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS) that allows relative 58
quantification of protein lactosylation and it is a reliable method to monitor the early Maillard 59
reaction as well as MRPs during milk processing (Meltretter, Becker, & Pischetsrieder, 2008). 60
The aim of the present paper, was to further improve the existing methodologies for the detection of 61
lysine and MRPs. A new method was designed which included direct hydrolysis along with stable 62
isotope dilution assay coupled with solid phase extraction and ion pairing liquid chromatography 63
tandem mass spectrometry (LC-MS/MS). The developed procedure allowed the simultaneous 64
detection of total lysine, furosine, CML and CEL. The method was tested on several foods: milk, 65
infant formulas, cookies, bread slices. The robustness after several injections and the reliability of 66
the results obtained were evaluated in soybean-based feed products obtained under severe thermal 67
treatment conditions. Data demonstrated satisfactory analytical performances on all tested samples 68
and results were perfectly in line with those previously obtained. 69
2. Material and methods 70
2.1 Chemicals and reagents 71
Acetonitrile, methanol and water for solid phase extraction (SPE) and LC-MS/MS determination 72
were obtained from Merck (Darmstadt, Germany). The ion pairing agent perfluoropentanoic acid, 73
trichloroacetic acid, hydrochloric acid (37%) and the analytical standards L-lysine hydrochloride 74
and [4,4,5,5-d4]-L-lysine hydrochloride (d4-Lys) were purchased from Sigma-Aldrich (St. Louis, 75
MO). Analytical standards Nε-(2-Furoylmethyl)-L-lysine (furosine), Nε-(Carboxymethyl)-L-lysine 76
(CML) and its respective deuterated standard Nε-(Carboxy[2H2]methyl)-L-Lysine (d2-CML) were 77
obtained from Polypeptide laboratories (Strasbourg, France), Nε-(Carboxyethyl)-L-lysine and its 78
internal standard Nε-(Carboxy[2H4]ethyl)-L-lysine (d4-CEL) were purchased from TRC-Chemicals 79
(North York, Canada). 80
2.2 Foods samples 81
Powdered infant formula and milk samples were purchased in a local market, biscuits samples and 82
bread slices were prepared according previous papers published by our group (Fiore, Troise, Mogol, 83
Roullier, Gourdon, Jian, et al., 2012; Vitaglione, Lumaga, Stanzione, Scalfi, & Fogliano, 2009). 84
UHT milk was prepared according to the procedure previously described (Troise, Fiore, 85
Colantuono, Kokkinidou, Peterson, & Fogliano, 2014). Raw milk (protein, 3.5%; fat, 1%) was 86
purchased in a local market. 87
2.2.1 Soybean samples 88
One batch of quartered raw soybeans was purchased from Rieder Asamhof GmbH & Co. KG 89
(Kissing, Germany).The raw soybeans were further processed at the hydrothermal cooking plant of 90
Amandus Kahl GmbH & Co. KG (Reinbeck, Germany). First, the beans were short-term 91
conditioned to reach a temperature of 80 °C after 45 seconds. Afterwards, the beans entered a 92
hydrothermic belt cooker at 72 °C and left inside for 3 min at a temperature of 70 °C. Then they 93
were expanded at 117 °C using an annular gap expander (Typ OEE 8, Amandus Kahl GmbH & Co. 94
KG, Reinbeck, Germany). The expanded soybeans were collected in a drying wagon for 10 min. 95
Then they were dried with air at 65 °C for 10 min and cooled for another 10 min to reach a final 96
moisture content of 12%. Afterwards, the expanded soybeans were autoclaved for 0, 5, 10, 15, 20, 97
25, 30, 35, 40, 45, 50 and 60 min at 110 °C and 1470 mbar using a fully controlled autoclave (Typ 98
HST 6x9x12, Zirbus Technology GmbH, Bad Grund, Germany). 99
2.3 Samples preparation 100
Lysine and its derivatives Nε-(2-Furoylmethyl)-L-lysine (furosine), Nε-(Carboxymethyl)-L-lysine 101
(CML), Nε-(Carboxyethyl)-L-lysine (CEL) were analyzed considering previous papers (Delatour, et 102
al., 2009; Fenaille, Parisod, Visani, Populaire, Tabet, & Guy, 2006; Troise, Dathan, Fiore, Roviello, 103
Di Fiore, Caira, et al., 2014) and introducing several modifications. Briefly, 100 mg of each sample 104
was accurately weighed in a screw capped flask with PTFE septa and 4 mL of hydrochloric acid (6 105
N) was added. The mixture was saturated by nitrogen (15 min at 2 bar) and hydrolyzed in an air 106
forced circulating oven (Memmert, Schwabach, Germany) for 20 h at 110° C. The mixture was 107
filtrated by polyvinylidene fluoride filters (PVDF, 0.22 Millipore, Billerica, MA) and 400 µl was 108
dried under nitrogen flow in order to prevent the oxidation of the constituents. The samples were 109
reconstituted in 370 µl of water and 10 µL of each internal standard d4-Lys, d2-CML and d4-CEL 110
was added in order to obtain a final concentration of 200 ng/mg of samples for both standards. 111
Samples were loaded onto equilibrated Oasis HLB 1 cc cartridges (Waters, Wexford, Ireland) and 112
eluted according to the method previously described, then 5 µl was injected onto the LC/MS/MS 113
system. 114
2.4 Liquid chromatography tandem mass spectrometry (LC-MS/MS) 115
Separation of furosine, CML, CEL, lysine and their respective internal standards was achieved on a 116
reversed – phase core shell HPLC column (Kinetex C18 2.6 µm, 2.1 mm x 100 mm, Phenomenex, 117
Torrance) using the following mobile phases: A, 5 mM perfluoropentanoic acid and B, acetonitrile 118
5 mM perfluoropentanoic acid. The compounds were eluted at 200 µL/min through the following 119
gradient of solvent B (t in [min]/[%B]): (0/10), (2/10), (5/70), (7/70), (9/90), (10/90), (12/10), 120
(15/10). Positive electrospray ionization was used for detection and the source parameters were 121
selected as follows: spray voltage: 5.0 kV; capillary temperature: 350 °C, dwell time 100 ms, cad 122
gas and curtain gas were set to 45 and 5 (arbitrary units). The chromatographic profile was recorded 123
in MRM mode and the characteristic transitions were monitored in order to improve selectivity 124
using an API 3000 triple quadrupole (ABSciex, Carlsbad, CA). All relevant parameters are 125
summarized in Table 1. 126
2.5 Analytical performances 127
CML, CEL, furosine and total lysine were quantified using a linear calibration curve built with 128
specific solutions of CML spiked with d2-CML, lysine and furosine spiked with d4-lysine and CEL 129
spiked with d4-CEL (final concentration of internal standards: 200 ng/ml) dissolved in water. The 130
limit of detection (LOD) and the limit of quantitation (LOQ) were monitored according to the signal 131
to noise ratio (Armbruster, Tillman, & Hubbs, 1994). The coefficients of determination r2 for the 4 132
analytes were tested plotting the ratio between the pure compounds and their respective internal and 133
the concentration of the pure compounds in the linearity range 5-1000 ng/mL. The internal standard 134
ratio was used for the quantification and the relative standard deviation of intraday and interday 135
assay was monitored three times each day and six times in different days. The recovery test was 136
monitored according to the concentration of the internal standards used and to the ratio between 137
labeled compounds and native compounds. 138
2.6 Statistical analysis 139
All of the analyses were performed in quadruplicate and the results expressed as mg/100 g of 140
protein. Statistical calculations were performed using Matlab R2009b (Natick, MA) while for mass 141
spectrometry data, Analyst version 1.4.2 (Applied Biosystems, Carlsbad, CA) was used. 142
3. Results and discussion 143
3.1 Liquid chromatography set up 144
Under the above described chromatographic conditions, typical retention time of CML and d2-CML 145
was 7.11 min, for d4-Lys and Lys it was 7.23 min, for furosine it was 7.91 min, while for CEL and 146
d4-CEL it was 7.36 min (Figure 2). Previous papers highlighted the problems due to the poor 147
retention of amino acids and their derived molecules on silica bonded and C-18 column (Frolov & 148
Hoffmann, 2008). Preliminary trials performed using C-18 column without the ion pairing agent 149
confirmed this feature: the retention was poor and the analytes co-eluted with the impurities on the 150
front of the chromatographic run with the consequent partial suppression of the signal associated to 151
the markers. Inadequate separation of the analytes was obtained also using polar end-capped 152
column; however a significant improvement was obtained using with this column 153
perfluoropentanoic acid as ion pairing agent. In these experimental conditions, the retention time 154
followed a typical reversed phase profile according to the polarity and to the steric hindrance of 155
each molecule, as previously observed by other papers published earlier (Fenaille, Parisod, Visani, 156
Populaire, Tabet, & Guy, 2006; Troise, Fiore, Roviello, Monti, & Fogliano, 2014). The presence of 157
the ion pairing agent charged the core shell residues increasing the retention and promoting the 158
selectivity of the positively charged CML, CEL, furosine, lysine and their respective internal 159
standards. The presence of a core shell phase increased of the resolution which directly reflects the 160
good performances of the reported method, the shape of the peak was maintained over each batch 161
and the retention time shift was lower than 0.5 min, highlighting the robustness of the analytical 162
performances. 163
3.2 Mass spectrometry set up 164
Mass spectrometry conditions were optimized by infusing singularly the seven standards directly in 165
the ion source. Collision energy, declustering potential, tube lens voltage along with spray voltage 166
and interface temperature were monitored in order to favor the formation of the typical 167
fragmentation pattern (Delatour, et al., 2009). The lysine derived compounds underwent the 168
formation of the fragment ion at 130 m/z which corresponds to the pipecolic acid generated by the 169
subsequent cyclization of the side chain of lysine and the loss of ε- amino group, similarly the mass 170
shift for deuterated standards d4-CEL and d4-Lys was +4 Da as consequence of the fragmentation 171
occurred on the side chain of lysine (Figure S1 in supplementary material section) (Yalcin & 172
Harrison, 1996). The MRM revealed the loss of formic acid giving the typical fragment at m/z 84; 173
the mass shift for the deuterated molecules was +4. The seven standards were also infused inside 174
the chromatographic flow in order to evaluate the interferences due to the ion pairing agent or to the 175
solvent and the results revealed that no enhancement or suppression effect can be ascribed to the 176
parameters monitored. 177
3.3 Analytical performances 178
The analytical performances of the method were tested against reproducibility, repeatability, limit 179
of detection (LOD), limit of quantification (LOQ), linearity, precision, carry-over and coefficient of 180
correlation (r2). Before and after each batch, three solutions of acetonitrile and water (90:10; 50:50 181
and 10:90) were injected in order to verify the absence of any contaminants with the same signal 182
and the same retention time of the analyzed molecules. The limit of detection and the limit of 183
quantitation were determined according to the procedure previously described. The concentration 184
0.1 ppb resulted in no signal, while the LOD was 0.5 ppb for CML and lysine while for CEL and 185
furosine it was 1 and 3 ppb, respectively. The slight differences among CML, CEL and furosine can 186
be related to the different stability in the injection conditions. By injecting these concentrations the 187
signal to noise ratio was always higher than 3. The LOQ were 5 ppb for CML, CEL and Lysine 188
while for furosine it was 9 ppb, as highlighted in Table 2. These values were perfectly in line with 189
those previously described for CML, CEL and lysine quantification by MS/MS (Delatour, et al., 190
2009; Tareke, Forslund, Lindh, Fahlgren, & Ostman, 2013) while for furosine the performance of 191
LOD and LOQ were below the values previously reported in milk (Bignardi, Cavazza, & Corradini, 192
2012). According to the LOD and LOQ, linearity was achieved in the range 5-1000 ppb for CML, 193
CEL and lysine, while for furosine the linearity range was between 9 and 1000 ppb. The carryover 194
effect was tested injecting after each point of the calibration curves a solution consisting in 195
acetonitrile and water (50:50, v/v) and verifying the absence of the target compounds. The linearity 196
of the calibration curves was evaluated three times in the same day (intraday assay for the 197
reproducibility) and three times for three subsequent days (interday assay for the repeatability) 198
using the ratio between the target compounds and their respective internal standard. The RSD (%) 199
among the three curves was always lower than 8%, demonstrating that external factors had marginal 200
impact on the performance of the method. Each point of the calibration curves was monitored using 201
two specific transitions: the most intensive fragment was used as quantifier, the lowest as qualifier. 202
For CML, CEL, furosine and lysine, the respective transitions of m/z 205−84.1, m/z 219.1−84.1, m/z 203
255.1−130.2, and m/z 147.2−130.2 were used as quantifier, whereas m/z 205−130.2, m/z 204
219.1−84.1, m/z 255.1−84, and m/z 147.2−84.1 were used as qualifier. CML was quantified using 205
d2-CML as internal standard (m/z 207−144.1 and 207−84 for quantification and confirmation, 206
respectively), CEL was quantified using d4-CEL (m/z 223−134.1 and 223 − 84 for quantification 207
and confirmation, respectively) whereas for furosine and lysine, d4-lysine was used (m/z 208
151.2−134.1 and m/z 151.1− 88 for quantification and confirmation, respectively). The use of d4-209
lysine as internal standard for the quantification and recovery of furosine was optimized by 210
monitoring the relative intensity of furosine standard towards d4-CEL, d2-CML and d4-lysine. A 211
mixture of the four standards (10 ppm) was directly infused in the ion source. Results revealed that 212
the intensity of the signal at m/z 151.2 and m/z 255 were similar and both were 15% higher than the 213
signal of d2-CML and d4-CEL. 214
The response of the method in food was tested during each batch evaluating the ratio between the 215
target compounds and the internal standard, these procedures confirmed and deepened the aspects 216
linked to the recovery assay: in each sample the ratio between the area of the analyte and the area of 217
the deuterated compounds was compared towards the calibration curve in order to obtain the final 218
concentration of the analytes in the matrix. The intensity of the internal standard in the samples and 219
in the standard was compared and the RSD (%) between the spiked samples and the spiked 220
standards was always lower than 10%. The recovery test was monitored in all the food matrix 221
according to the intensity of the internal standard, the results were 91.1 ± 8.4, 84.2 ± 7.4, 88.0 ± 6.9 222
for d2-CML, d4-CEL and d4-Lysine. 223
3.4 CML, CEL, furosine and total Lysine in food 224
Powdered samples were freeze dried prior analysis in order to remove the interferences due to the 225
humidity. The extraction procedure of MRPs is characterized by three key steps: the reduction with 226
sodium borohydride, the hydrolysis with hydrochloric acid and the stable isotope dilution assay 227
prior ion pairing solid phase extraction. According to the nature of protein and to their concentration 228
each of the above listed can influence the yield and the efficiency of the extraction. The reduction 229
with sodium borohydride promotes the conversion of free fructosyl-lysine into hexitol-lysine in 230
order to avoid the overestimation of CML, CEL (Niquet-Leridon & Tessier, 2011). Moreover, the 231
use of sodium borohydride is recommended when the concentration of free unstable Amadori 232
products is high. Unfortunately, the use of this reducing agent had several drawbacks: protein 233
degradation and free counterpart losses during the reduction, precipitation and purification 234
procedure; moreover, the use of sodium borohydride can interfere with the release of furosine with 235
the above mentioned reduction of fructosyl-lysine into hexitol-lysine. After several preliminary 236
measurements it was decided to avoid the reduction. A good compromise between the detection of 237
furosine and that of CML/CEL was achieved controlling the oxidation under nitrogen. In particular, 238
prior the acidic hydrolysis the screw capped flasks were saturated with nitrogen in order to reduce 239
the effect of autoxidation and control the reaction pathway (Yaylayan & Huyghuesdespointes, 240
1994). 241
The use of hydrochloric acid is a mandatory step for the hydrolysis of peptide bonds and for the 242
release of amino acids, MRPs and for the conversion of fructosyl-lysine into furosine. Different 243
concentrations of protein per mL of hydrochloric acid can lead to different efficiency of the 244
hydrolysis with the consequent underestimation of lysine content. In the present study, the 245
extraction procedure was optimized in order to promote the dehydration reaction that leads to the 246
formation of furosine and to the release of MRPs (Krause, Knoll, & Henle, 2003; Mossine & 247
Mawhinney, 2007). Further studies will be conducted in order to compare the effect of time and 248
concentration of hydrochloric acid on lysine release, mainly in protein rich samples. 249
The above described analytical performances were tested in food and feed samples in order to 250
verify the robustness of the method. Several thermally treated foods were tested: powdered infant 251
formula, low lactose milk, lab scale UHT milk, biscuits samples, bread (all prepared according to 252
three different procedures previously described by our group) and powdered soybean-based feed 253
products (prepared at industry scale). All data are summarized in Table 3. The concentration of 254
CML in powdered infant formula analyzed ranged from 8.22 ± 0.31 mg/100 g of protein to 14.81 ± 255
0.92 mg/100 g of protein, while CEL and furosine ranged from 0.71 ± 0.02 mg/100 g of protein to 256
1.31 ± 0.11 mg/100 g of protein and 471.9± 22.3 mg/100 g of protein to 639.4± 21.1 mg/100 g of 257
protein, respectively. The concentration of total lysine varied from 9.89 ± 0.88 to 13.12 ± 0.78 % of 258
total protein. In low lactose milk the content of lysine was 5.21± 0.30 g/100 g of protein, while the 259
concentration of CEL and furosine was 0.28 mg ± 0.01 mg/100 g of protein and 12.32 ± 0.31 260
mg/100 g of protein, respectively. CML was 1.28 mg ± 0.11 mg/100 g of protein and this value was 261
perfectly in line with the one previously obtained. Lab scale UHT milk was prepared in order to 262
verify the effect on raw cow milk; while the lysine content was of the same order of magnitude of 263
the low lactose milk (4.71 ± 0.22 mg/100 g of protein), the concentration of the three markers of the 264
MR was 18.41 ± 0.93, 1.12 ± 0.02 and 14.41 ± 1.02 mg/100 g of protein for CML, CEL and 265
furosine respectively. The results obtained were perfectly in line with those previously obtained for 266
the three categories of milk (Fenaille, Parisod, Visani, Populaire, Tabet, & Guy, 2006; Tareke, 267
Forslund, Lindh, Fahlgren, & Ostman, 2013), specifically the CML in low lactose milk was similar 268
to one previously obtained by our group for LC-MS/MS analysis (Troise, et al., 2014). The 269
concentration of CML and furosine was closed to the range previously obtained: 2.2 – 30.8 and 0.8 270
– 3.7 mg/100 g of protein for furosine and CML, respectively (de Sereys, Muller, Desic, Troise, 271
Fogliano, Acharid, et al., 2014) . 272
In bakery products CML content was 43.75 ± 2.02 and 27.15 ± 0.61 mg/100 g of protein for biscuits 273
samples and bread slices, respectively, while CEL and furosine were 46.25 ± 3.01 and 10.01 ± 274
0.61 and 10.91 ± 0.01 and 98.55 ± 4.61 mg/100 g of protein for biscuits and bread, respectively. 275
The lysine content was almost similar in the two products: 5.01 ± 0.04 and 5.81 ± 0.04 g/100 g of 276
protein, even if the protein content was 6% and 8% for biscuits and bread. The results here reported 277
were of the same order of magnitude as the ones previously reported. Hull et al., analyzed several 278
kinds of bread and other bakery products and the concentration of CML ranged from 2.6 to 45.1 279
mg/100 g of protein for wheaten bread and potato bread, respectively (Hull, Woodside, Ames, & 280
Cuskelly, 2012). On the other hand He and coworker reported higher values for wholemeal bread: 281
CML ranged from 66.72 to 109.9 mg/100g of protein and CEL ranged from 53.30 to 82.04 mg/100 282
g protein for bread, while in biscuits samples the concentrations varied from 50.8 to 116.7 and 283
15.87 to 45.26 mg/100g protein for CML and CEL, respectively (He, Zeng, Zheng, He, & Chen, 284
2014). Interestingly, the concentration of furosine in bread (after 20 min at 200° C) is similar to the 285
one reported by Capuano and coworker: after 13 min the concentration of furosine increased up to 286
200 mg/100 g of protein and it quickly decreased up to 20 mg/100 g protein at the end of the 287
thermal treatment (Capuano, Ferrigno, Acampa, Ait-Ameur, & Fogliano, 2008). A similar kinetic 288
profile was observed also by Ramirez-Jimenez and coworker in sliced bread: the concentration of 289
furosine at the end of the process was 79.3 mg/100 g of protein while after 12 min it reached a 290
concentration higher than 200 mg/100 g protein (Ramirez-Jimenez, Garcia-Villanova, & Guerra-291
Hernandez, 2001). In biscuit samples the kinetic profile revealed similar trends to the ones obtained 292
for bread; as a consequence at the end of the thermal process the concentration of furosine value of 293
10.01 ± 0.61 mg/100 g of protein was comparable to those of sucrose-containing cookies reported 294
by previous authors (Gökmen, Serpen, Açar, & Morales, 2008). 295
The above described analytical performances were evaluated in industrially prepared soybean feeds 296
in order to verify the main advantages of the method on industrial sampling. The simultaneous 297
quantification of the four analytes allowed a direct overview of the extent of the MR, where the 298
concentration of lysine and the formation of furosine, CEL and CML can be easily related to the 299
final quality of foods using a single extraction and a single injection. According to the procedure 300
described in material and methods section, soybeans were incubated at 110° C for one hour in an 301
autoclave and the kinetic profile was reported in Figure 3. The initial concentration of lysine was 302
3.45 ± 0.12 g/100 g of protein while CML, CEL and furosine were 9.94 ± 0.74, 0.98 ± 0.04 and 303
24.24 ± 1.74 mg/100 g of protein respectively. After 30 minutes the concentration of furosine 304
reached the highest values: 108.01 ± 8.97, then it rapidly decreased up to 60.58 ± 3.75 mg/100 g of 305
protein after 55 min. According to the reaction mechanism the degradation of the Amadori products 306
was followed by the increase of CML: at the end of the thermal treatment its concentration was 307
higher than 76 mg/100 g of protein. CEL reached the maximum concentration after 45 minutes 308
(2.41 ± 0.24 mg/100 g of protein), then it decreased probably due to degradation processes or to the 309
blockage of methylglyoxal by other compounds. The degradation of lysine was constant throughout 310
the thermal treatment, after 60 min lysine concentration was 2.60 ± 0.08 g/100 g of protein thus 311
around 23%. Several studies reported the effect of soy proteins in the development of the MR 312
focusing on soy health benefits and on the presence of functional molecules able to control the 313
extent of the MR (Palermo, Fiore, & Fogliano, 2012). 314
This paper represents the first example of a systematic study on the relationship between thermal 315
treatments, MR and soybean products in feeds and in pet food a topic recently attracting the 316
attention of the scientific community. In fact, it has been observed that the average daily intake 317
(mg/kg body weight0.75
) of HMF is 122 times higher for dogs and 38 times higher for cats than 318
average intake for adult humans. Possible health risks, such diabetes and renal failure, can be 319
associated to the intake of MRPs not only in human, but also in pets (van Rooijen, Bosch, van der 320
Poel, Wierenga, Alexander, & Hendriks, 2013). 321
4. Conclusion 322
The analytical method allowed a comprehensive approach in the analysis of MRPs, simultaneously 323
determining both lysine and its heat-induced derivatives. Up to now the golden standards for MRPs 324
detection were RP-HPLC with UVvis detection for furosine and LC-MS/MS for CML, CEL and 325
lysine, respectively. These results showed that the extraction procedure with nitrogen and 326
hydrochloric acid provided a good compromise for the simultaneous detection of the four analytes. 327
The analytical performances showed high sensitivity and good reproducibility and repeatability in 328
several foods. Quantitative data were fully in line with those previously obtained by other authors 329
on similar foods. The simultaneous detection of the four analytes offered a sensitive tool for the 330
kinetic modeling on neoformed contaminant reaction routes monitoring the precursor lysine, the 331
intermediate furosine via the indirect analysis of the Amadori products and the end-products CEL 332
and CML. The simultaneous monitoring of all compounds allowed to minimize the variability 333
among different samples and to combine the reaction steps starting from lysine blockage, Amadori 334
compounds formation and fragmentation, CML and CEL formation. 335
The research of Antonio Dario Troise at Wageningen University was partly supported by UniNA 336
and Compagnia di San Paolo, in the frame of Programme STAR. 337
338
The authors declare no conflict of interests. 339
References 340
Ahmed, M. U., Thorpe, S. R., & Baynes, J. W. (1986). Identification of N-Epsilon-Carboxymethyllysine as a 341
Degradation Product of Fructoselysine in Glycated Protein. Journal of Biological Chemistry, 261(11), 4889-342
4894. 343
Armbruster, D. A., Tillman, M. D., & Hubbs, L. M. (1994). Limit of Detection (Lod) Limit of Quantitation (Loq) - 344
Comparison of the Empirical and the Statistical, Methods Exemplified with Gc-Ms Assays of Abused Drugs. 345
Clinical Chemistry, 40(7), 1233-1238. 346
Bignardi, C., Cavazza, A., & Corradini, C. (2012). Determination of furosine in food products by capillary zone 347
electrophoresis-tandem mass spectrometry. Electrophoresis, 33(15), 2382-2389. 348
Brownlow, S., Cabral, J. H. M., Cooper, R., Flower, D. R., Yewdall, S. J., Polikarpov, I., North, A. C., & Sawyer, L. 349
(1997). Bovine< i> β</i>-lactoglobulin at 1.8 Å resolution—still an enigmatic lipocalin. Structure, 5(4), 481-350
495. 351
Capuano, E., Ferrigno, A., Acampa, I., Ait-Ameur, L., & Fogliano, V. (2008). Characterization of the Maillard reaction 352
in bread crisps. European Food Research and Technology, 228(2), 311-319. 353
Charissou, A., Ait-Ameur, L., & Birlouez-Aragon, I. (2007). Evaluation of a gas chromatography/mass spectrometry 354
method for the quantification of carboxymethyllysine in food samples. Journal of Chromatography A, 1140(1-355
2), 189-194. 356
de Sereys, A. L., Muller, S., Desic, S., Troise, A., Fogliano, V., Acharid, A., Lacotte, P., & Birlouez-Aragon, I. (2014). 357
Potential of the FAST index to characterize infant formula quality. In V. Preedy (Ed.), Handbook ofdietary and 358
nutritional aspects of bottle feeding ): Wageningen Academic Publishers. 359
Delatour, T., Hegele, J., Parisod, V., Richoz, J., Maurer, S., Steven, M., & Buetler, T. (2009). Analysis of advanced 360
glycation endproducts in dairy products by isotope dilution liquid chromatography-electrospray tandem mass 361
spectrometry. The particular case of carboxymethyllysine. Journal of Chromatography A, 1216(12), 2371-362
2381. 363
Erbersdobler, H. F., & Somoza, V. (2007). Forty years of furosine - Forty years of using Maillard reaction products as 364
indicators of the nutritional quality of foods. Molecular Nutrition & Food Research, 51(4), 423-430. 365
Fenaille, F., Parisod, V., Visani, P., Populaire, S., Tabet, J. C., & Guy, P. A. (2006). Modifications of milk constituents 366
during processing: A preliminary benchmarking study. International Dairy Journal, 16(7), 728-739. 367
Fiore, A., Troise, A. D., Mogol, B. A., Roullier, V., Gourdon, A., Jian, S. E., Hamzalioglu, B. A., Gokmen, V., & 368
Fogliano, V. (2012). Controlling the Maillard Reaction by Reactant Encapsulation: Sodium Chloride in 369
Cookies. Journal of Agricultural and Food Chemistry, 60(43), 10808-10814. 370
Frolov, A., & Hoffmann, R. (2008). Separation of Amadori peptides from their unmodified analogs by ion-pairing RP-371
HPLC with heptafluorobutyric acid as ion-pair reagent. Analytical and Bioanalytical Chemistry, 392(6), 1209-372
1214. 373
Gökmen, V., Serpen, A., Açar, Ö. Ç., & Morales, F. J. (2008). Significance of furosine as heat-induced marker in 374
cookies. Journal of cereal science, 48(3), 843-847. 375
Hartkopf, J., Pahlke, C., Ludemann, G., & Erbersdobler, H. F. (1994). Determination of N-Epsilon-376
Carboxymethyllysine by a Reversed-Phase High-Performance Liquid-Chromatography Method. Journal of 377
Chromatography A, 672(1-2), 242-246. 378
He, J. L., Zeng, M. M., Zheng, Z. P., He, Z. Y., & Chen, J. (2014). Simultaneous determination of N (epsilon)-379
(carboxymethyl) lysine and N (epsilon)-(carboxyethyl) lysine in cereal foods by LC-MS/MS. European Food 380
Research and Technology, 238(3), 367-374. 381
Henle, T., Zehetner, G., & Klostermeyer, H. (1995). Fast and Sensitive Determination of Furosine. Zeitschrift Fur 382
Lebensmittel-Untersuchung Und-Forschung, 200(3), 235-237. 383
Hidalgo, F. J., & Zamora, R. (2005). Interplay between the Maillard reaction and lipid peroxidation in biochemical 384
systems. Maillard Reaction: Chemistry at the Interface of Nutrition, Aging, and Disease, 1043, 319-326. 385
Hodge, J. E. (1953). Chemistry of Browning Reactions in Model Systems. J. Agric. Food Chem.(1), 928–943. 386
Hull, G. L. J., Woodside, J. V., Ames, J. M., & Cuskelly, G. J. (2012). N-epsilon-(carboxymethyl)lysine content of 387
foods commonly consumed in a Western style diet. Food Chemistry, 131(1), 170-174. 388
Krause, R., Knoll, K., & Henle, T. (2003). Studies on the formation of furosine and pyridosine during acid hydrolysis of 389
different Amadori products of lysine. European Food Research and Technology, 216(4), 277-283. 390
Leclere, J., Birlouez-Aragon, I., & Meli, M. (2002). Fortification of milk with iron-ascorbate promotes lysine glycation 391
and tryptophan oxidation. Food Chemistry, 76(4), 491-499. 392
Meltretter, J., Becker, C. M., & Pischetsrieder, M. (2008). Identification and site-specific relative quantification of beta-393
lactoglobulin modifications in heated milk and dairy products. Journal of Agricultural and Food Chemistry, 394
56(13), 5165-5171. 395
Mossine, V. V., & Mawhinney, T. P. (2007). N-alpha-(1-DeOXY-D-fructos-1-yl)-L-histidine ("D-fructose-L-396
histidine"): a potent copper chelator from tomato powder. Journal of Agricultural and Food Chemistry, 55(25), 397
10373-10381. 398
Nguyen, H. T., van der Fels-Klerx, H. J., & van Boekel, M. A. J. S. (2013). N ϵ-(carboxymethyl)lysine: A Review on 399
Analytical Methods, Formation, and Occurrence in Processed Food, and Health Impact. Food Reviews 400
International, 30(1), 36-52. 401
Niquet-Leridon, C., & Tessier, F. J. (2011). Quantification of N-epsilon-carboxymethyl-lysine in selected chocolate-402
flavoured drink mixes using high-performance liquid chromatography-linear ion trap tandem mass 403
spectrometry. Food Chemistry, 126(2), 655-663. 404
Palermo, M., Fiore, A., & Fogliano, V. (2012). Okara promoted acrylamide and carboxymethyl-lysine formation in 405
bakery products. J Agric Food Chem, 60(40), 10141-10146. 406
Pischetsrieder, M., & Henle, T. (2012). Glycation products in infant formulas: chemical, analytical and physiological 407
aspects. Amino Acids, 42(4), 1111-1118. 408
Ramirez-Jimenez, A., Garcia-Villanova, B., & Guerra-Hernandez, E. (2001). Effect of toasting time on the browning of 409
sliced bread. Journal of the Science of Food and Agriculture, 81(5), 513-518. 410
Tareke, E., Forslund, A., Lindh, C. H., Fahlgren, C., & Ostman, E. (2013). Isotope dilution ESI-LC-MS/MS for 411
quantification of free and total N epsilon-(1-Carboxymethyl)-L-Lysine and free N epsilon-(1-Carboxyethyl)-L-412
Lysine: Comparison of total N epsilon-(1-Carboxymethyl)-L-Lysine levels measured with new method to 413
ELISA assay in gruel samples. Food Chemistry, 141(4), 4253-4259. 414
Technische Universität Dresden. (2014). AGE Database. In, vol. 2014). 415
Tessier, F. J., & Birlouez-Aragon, I. (2012). Health effects of dietary Maillard reaction products: the results of ICARE 416
and other studies. Amino Acids, 42(4), 1119-1131. 417
Troise, A. D., Dathan, N. A., Fiore, A., Roviello, G., Di Fiore, A., Caira, S., Cuollo, M., De Simone, G., Fogliano, V., 418
& Monti, S. M. (2014). Faox enzymes inhibited Maillard reaction development during storage both in protein 419
glucose model system and low lactose UHT milk. Amino Acids, 46(2), 279-288. 420
Troise, A. D., Fiore, A., Colantuono, A., Kokkinidou, S., Peterson, D. G., & Fogliano, V. (2014). Effect of olive mill 421
wastewater phenol compounds on reactive carbonyl species and maillard reaction end-products in ultrahigh-422
temperature-treated milk. J Agric Food Chem, 62(41), 10092-10100. 423
Troise, A. D., Fiore, A., Roviello, G., Monti, S. M., & Fogliano, V. (2014). Simultaneous quantification of amino acids 424
and Amadori products in foods through ion-pairing liquid chromatography-high-resolution mass spectrometry. 425
Amino Acids. 426
Vallejo-Cordoba, B., Mazorra-Manzano, M. A., & Gonzalez-Cordova, A. F. (2004). New capillary electrophoresis 427
method for the determination of furosine in dairy products. Journal of Agricultural and Food Chemistry, 428
52(19), 5787-5790. 429
van Boekel, M., Fogliano, V., Pellegrini, N., Stanton, C., Scholz, G., Lalljie, S., Somoza, V., Knorr, D., Jasti, P. R., & 430
Eisenbrand, G. (2010). A review on the beneficial aspects of food processing. Molecular Nutrition & Food 431
Research, 54(9), 1215-1247. 432
van Rooijen, C., Bosch, G., van der Poel, A. F. B., Wierenga, P. A., Alexander, L., & Hendriks, W. H. (2013). The 433
Maillard reaction and pet food processing: effects on nutritive value and pet health. Nutrition research reviews, 434
26(02), 130-148. 435
Vitaglione, P., Lumaga, R. B., Stanzione, A., Scalfi, L., & Fogliano, V. (2009). beta-Glucan-enriched bread reduces 436
energy intake and modifies plasma ghrelin and peptide YY concentrations in the short term. Appetite, 53(3), 437
338-344. 438
Yalcin, T., & Harrison, A. G. (1996). Ion chemistry of protonated lysine derivatives. Journal of mass spectrometry, 439
31(11), 1237-1243. 440
Yaylayan, V. A., & Huyghuesdespointes, A. (1994). Chemistry of Amadori Rearrangement Products - Analysis, 441
Synthesis, Kinetics, Reactions, and Spectroscopic Properties. Critical Reviews in Food Science and Nutrition, 442
34(4), 321-369. 443
444
445
446
447
448
449
450
451
452
453
Figure legend 454
Figure 1: Effect of glucose and dicarbonyls on the formation of protein-bound MRPs. At the 455
bottom the structure of β-Lactoglobulin (Brownlow, Cabral, Cooper, Flower, Yewdall, Polikarpov, 456
et al., 1997). 457
Figure 2: Extracted ion chromatogram of the four target molecules and their respective internal 458
standards 459
Figure 3: Kinetic profile of the precursor lysine (green), intermediate, furosine (blue) and end-460
products, CML and CEL (red). 461
Figure S1: Fragmentation pathway for lysine and its deuterated internal standard d4-lysine. The 462
structures of pipecolic acid and 1,2,3,4‐tetrahydropyridin‐1‐ium ion was reported (Yalcin & 463
Harrison, 1996). 464
465
Table legend 466
Table 1: Mass spectrometry set up 467
Table 2: Analytical performances for the four analytes and their respective internal standards 468
Table 3: MRPs concentration after 8 replicates in different samples, the results for CML, CEL and 469
furosine were reported as mg/100 g of protein, except for lysine. The results were compared to the 470
AGE Database (Technische Universität Dresden, 2014). 471
Tables
Table 1
Compounds [M+H]+ Fragments CE (V) DP (V)
CML 205 84 29 30
130.2 27 30
d2 -CML 207 84 30 20
144 21 20
130 17 20
Furosine 255.1 130 18 21
84.4 28 21
Lys 147.2 130.2 16 30
84.1 24 30
d4 –Lys 151.3 134.1 15 30
88.2 26 30
CEL 219.2 130.3 20 30
84.0 28 30
d4-CEL 223 134.1 18 25
88.0 30 25
Table 1: Mass spectrometry set up
Table 2
Compound LOD LOQ RSD [%] Linearity range r2 Recovery
CML 0.5 ppb 5 ppb 7 5-1000 ng/ml > 0.99 91.1 ± 8.4
CEL 1 ppb 5 ppb 5 5-1000 ng/ml > 0.99 84.2 ± 7.4
Lysine 0.5 ppb 5 ppb 5 5-1000 ng/ml > 0.99 88 .0± 6.9
Furosine 3 ppb 9 ppb 8 9-1000 ng/ml > 0.99 88.0 ± 6.9
Table 2: Analytical performances for the four analytes and their respective internal standards
Table 3:
Food CML CEL Furosine Lysine (g/100 g protein)
Infant formula -1 8.22 ± 0.31 0.71 ± 0.02 471.91 ± 22.31 9.89 ± 0.88
Infant formula -2 10.4 ± 0.52 0.85 ± 0.06 542.53 ± 11.91 12.24 ±0.91
Infant formula -3 10.9 ± 1.03 1.10 ± 0.05 574.5 ± 44.12 13.12 ± 0.78
Infant formula -4 14.81 ± 0.92 1.31 ± 0.11 639.4 ± 21.11 10.28 ± 1.01
Age Database 0.6 – 40.5 / Up to 1819 /
Low lactose milk 1.28 ± 0.11 0.28 ± 0.01 12.32 ± 0.31 5.21 ± 0.30
Age Database 1.4 / / /
Lab scale UHT milk 18.41 ± 0.93 1.12 ± 0.02 14.41 ±1.02 4.71 ± 0.22
Age Database 0.9-8.3 / 12.4 – 220.0 /
Biscuits 43.75 ± 2.02 46.25 ± 3.01 10.01 ± 0.61 5.01 ± 0.04
Bread slices 27.15 ± 0.61 10.91 ± 0.01 98.55 ± 4.61 5.81 ± 0.04
Age Database 2.6 – 45.1 / / /
Table 3: MRPs concentration after 8 replicates in different samples, the results for CML, CEL and
furosine were reported as mg/100 g of protein, except for lysine. The results were compared to the
AGE Database (Technische Universität Dresden, 2014).
Tandem mass spectrometry and stable isotope dilution ensured reliable performances.
The method achieved simultaneous detection of CML, CEL, Lysine and furosine.
CML, CEL, Lysine and furosine were quantified in several foods.
The analysis of the four markers paved the way for a better quality control.